Not an issue, per-se, but definitely something we should add.
Currently, in the literature, most academics report the uncertainty in parameter i of a least squares fit as sqrt(cov[i][i]). Given the large covariances in equation of state fitting, this is a severe misuse of statistics. I would like to see something better, for example, for EoS fitting:
Visualisation for the covariances
Confidence intervals in volume at a given P, T (using the delta method?)
bobmyhill
commented
7 years ago
Not an issue, per-se, but definitely something we should add.
Currently, in the literature, most academics report the uncertainty in parameter i of a least squares fit as sqrt(cov[i][i]). Given the large covariances in equation of state fitting, this is a severe misuse of statistics. I would like to see something better, for example, for EoS fitting: